CN-122021333-A - Oil gas development favorable region prediction method based on Winsorized particle size mean curve simulation

CN122021333ACN 122021333 ACN122021333 ACN 122021333ACN-122021333-A

Abstract

The invention discloses an oil gas development favorable region prediction method based on Winsorized particle size mean curve simulation, which comprises the following steps of processing rock particle size data by adopting Winsorized algorithm based on laser particle size experimental data, removing abnormal interference values, simulating a well hole particle size mean longitudinal curve by inputting a well logging response parameter by means of a XLSTM-transducer mixed neural network model, optimizing shear modulus density and longitudinal and transverse wave speed ratio as sensitive attribute parameters of particle size mean by combining seismic data, forming two sets of three-dimensional attribute data bodies by combining waveform structure inversion, and finally forming Winsorized particle size mean inversion body by carrying out model training on nonlinear relation between particle size mean and seismic sensitive attribute parameters based on a machine learning method, so that the spatial distribution rule of a relatively coarse particle size reservoir can be defined, and effective prediction of a development favorable region is realized.

Inventors

ZHANG CHONG
YIN LULU
ZHOU WEI
MENG DI
CHEN RUXIAN
ZHANG QIAN
ZHA YUQIANG
XU XIAOTING
GUAN YAO
ZHANG ZHU

Assignees

中海石油(中国)有限公司海南分公司

Dates

Publication Date: 20260512
Application Date: 20260206

Claims (10)

1. The oil gas development favorable region prediction method based on Winsorized particle diameter mean curve simulation is characterized by comprising the following steps of, S1, performing laser granularity test analysis based on a wall core sample of a well drilling hole to obtain rock granularity distribution data of different depths of the well drilling; S2, adopting Winsorized mean algorithm to process the particle size data, and calculating the mean particle size of the abnormal value interference resistance; s3, establishing Winsorized particle size average value sample database; S4, constructing XLSTM-transducer mixed neural network model, taking a conventional logging curve as input, and taking the average value of the particle size of Winsorized as output for training to obtain a prediction model capable of simulating a continuous borehole longitudinal Winsorized particle size average value curve; S5, selecting a geophysical elastic parameter shear modulus density mu rho sensitive to a Winsorized particle size average value and a longitudinal and transverse wave speed ratio Vp/Vs, combining a prestack three-dimensional seismic data body, and respectively forming two elastic parameter inversion three-dimensional attribute data bodies closely related to the Winsorized particle size average value through waveform structure inversion; And S6, establishing a nonlinear correlation training model of the Winsorized particle size mean curve, the shear modulus density muρ and the longitudinal and transverse wave speed ratio Vp/Vs, simultaneously writing an inversion body of the shear modulus density muρ and the longitudinal and transverse wave speed ratio Vp/Vs into the training model to form a Winsorized particle size mean inversion body, and determining a spatial distribution rule of a relatively coarse particle size reservoir through the Winsorized particle size mean inversion body to realize effective prediction of a development beneficial region.
2. The method for predicting an oil and gas development favorable region based on Winsorized particle size mean curve simulation according to claim 1, wherein in S1, the measurement range of the laser particle size test is 0.02-2000 μm.
3. The method for predicting the favorable oil and gas development area based on Winsorized particle size mean curve simulation according to claim 1 or 2, wherein the step S2 comprises the following steps, S21, determining a data boundary of the Windsor treatment, and setting a minimum threshold value a and a maximum threshold value b; s22, performing Windsor treatment on the original particle size sample data; s23, calculating Winsorized particle size average values based on the data set after the Windsor.
4. The method for predicting an oil and gas development favorable region based on Winsorized-particle diameter mean curve simulation according to claim 3, wherein in S21, the extremely small threshold value a=d (k+1) and the extremely large threshold value b=d (k-1) are determined by quantile truncation of ordered sample data { d1, d2,..dw, dn }, and d1< d2<, < dn >, wherein the truncated extreme value proportion accounts for 5% to 15% of the total sample data, a is an extremely small threshold value, b is an extremely large threshold value, k is a boundary data point, d is particle diameter, and different particle diameter particle ratios are obtained by laser particle size experiments.
5. The method for predicting an oil and gas development favorable region based on Winsorized particle size mean curve simulation as set forth in claim 4, wherein in S22, the data Windsor processing formula is as follows: , Where di is the i-th data point in the raw particle size sample data.
6. The method for predicting an oil and gas development favorable region based on Winsorized particle size mean curve simulation according to claim 4 or 5, wherein in S23, the Winsorized particle size mean calculation formula is as follows: ; wherein n is the total number of samples; the sample value is the i-th sample value after the warm-salvation, if the sample is smaller than a, the sample value is replaced by a, and if the sample is larger than b, the sample value is replaced by b; Is Winsorized mean.
7. The method for predicting an oil and gas development favorable region based on Winsorized particle size mean curve simulation according to claim 1 or 2, wherein in the step S4, the conventional logging curve comprises at least one of a natural gamma curve, a neutron porosity curve and a density curve, and a XLSTM network in the XLSTM-Transformer hybrid neural network model is used for capturing local trend and periodic change of the vertical particle size mean, and a Transformer network is used for capturing long-distance dependence in a stratum sequence.
8. The method for predicting an oil and gas development favorable region based on Winsorized particle size mean curve simulation according to claim 1 or 2, wherein in the S5, the shear modulus density μρ and the longitudinal and transverse wave velocity ratio Vp/Vs with the strongest correlation with Winsorized particle size mean are screened out from a plurality of pre-stack seismic elasticity parameters by adopting a method combining Theil-Sen regression and Spearman correlation analysis.
9. The method for predicting the favorable region of oil gas development based on Winsorized particle size mean curve simulation according to claim 1 or 2, wherein in the step S6, the training model is established by using an artificial neural network learning method based on Petrel software, and the artificial neural network learning method is a self-learning process for establishing a nonlinear mapping relation between the Winsorized particle size mean and the shear modulus density μρ and longitudinal/transverse wave velocity ratio Vp/Vs through error back propagation and gradient optimization.
10. A hydrocarbon development favorable region prediction system based on Winsorized particle size mean curve simulation is characterized in that the hydrocarbon development favorable region prediction method based on Winsorized particle size mean curve simulation is operated according to any one of claims 1 to 9.

Description

Oil gas development favorable region prediction method based on Winsorized particle size mean curve simulation Technical Field The invention belongs to the technical field of oil and gas field development, and particularly relates to a method for predicting an oil and gas development favorable region based on Winsorized particle size mean curve simulation. Background In the offshore strong heterogeneous gas reservoir development process, accurate prediction of development of a favorable region is critical to well site deployment of gas field development. The traditional method mainly depends on a logging curve and seismic data, but for a hypotonic reservoir layer with physical properties controlled by granularity, the conventional logging response characteristics are not obvious, so that the prediction precision of a relatively hypertonic reservoir layer is insufficient. In the prior art, particle size analysis is mostly limited to laboratory scale, continuous prediction between wells is difficult to realize, and an effective multi-scale data fusion method is lacked. Thus, there is a need for an innovative prediction method that can integrate core experiments, well logging, and seismic data. Disclosure of Invention The invention aims to provide a method for predicting an oil gas development favorable region based on Winsorized particle size mean curve simulation, which solves the problems of low prediction precision and poor data continuity in the prior art. In order to solve the technical problems, the invention adopts the technical proposal that the method for predicting the oil gas development favorable region based on Winsorized particle diameter mean curve simulation comprises the following steps, S1, performing laser granularity test analysis based on a wall core sample of a well drilling hole to obtain rock granularity distribution data of different depths of the well drilling; S2, adopting Winsorized mean algorithm to process the particle size data, and calculating the mean particle size of the abnormal value interference resistance; s3, establishing Winsorized particle size average value sample database; S4, constructing XLSTM-transducer mixed neural network model, taking a conventional logging curve as input, and taking the average value of the particle size of Winsorized as output for training to obtain a prediction model capable of simulating a continuous borehole longitudinal Winsorized particle size average value curve; S5, selecting a geophysical elastic parameter shear modulus density mu rho sensitive to a Winsorized particle size average value and a longitudinal and transverse wave speed ratio Vp/Vs, combining a prestack three-dimensional seismic data body, and respectively forming two elastic parameter inversion three-dimensional attribute data bodies closely related to the Winsorized particle size average value through waveform structure inversion; And S6, establishing a nonlinear correlation training model of the Winsorized particle size mean curve, the shear modulus density muρ and the longitudinal and transverse wave speed ratio Vp/Vs, simultaneously writing an inversion body of the shear modulus density muρ and the longitudinal and transverse wave speed ratio Vp/Vs into the training model to form a Winsorized particle size mean inversion body, and determining a spatial distribution rule of a relatively coarse particle size reservoir through the Winsorized particle size mean inversion body to realize effective prediction of a development beneficial region. Further, in the step S1, the measurement range of the laser granularity test is 0.02 μm to 2000 μm. Further, the step S2 comprises the following steps, S21, determining a data boundary of the Windsor treatment, and setting a minimum threshold value a and a maximum threshold value b; s22, performing Windsor treatment on the original particle size sample data; s23, calculating Winsorized particle size average values based on the data set after the Windsor. Further, in S21, the minimum threshold value a=d (k+1) and the maximum threshold value b=d (k-1) are determined by quantile truncation of the sequenced sample data { d1, d2,..dw }, and d1< d2< dn >, wherein the truncated extreme value proportion accounts for 5% to 15% of the total sample data, a is the minimum threshold value, b is the maximum threshold value, k is the boundary data point, d is the particle size, and the laser particle size experiment obtains the particle duty of different particle sizes. Further, in the step S22, the data Windsor processing formula is as follows: , Where di is the i-th data point in the raw particle size sample data. Further, in the step S23, the calculation formula of the average particle diameter of Winsorized is as follows:; wherein n is the total number of samples; the sample value is the i-th sample value after the warm-salvation, if the sample is smaller than a, the sample value is replaced by a, and if the sample is larger than b, the sample